Event-driven research lab

Infrastructure and alpha, co‑evolving.

We forecast the minutes after an event and trade them at low latency, on infrastructure that rewrites itself as its models decay.

Event-driven

Low latency

Agentic

Self-evolving

01 Loop

The loop is the product.

Agents run every step. Capital and risk stay with us.

01

Observe

One ordered stream of markets and news.

02

Hypothesise

Candidates are generated inside stated bounds.

03

Experiment

Runs burst onto ephemeral compute.

04

Evaluate

Out-of-sample decides, and most candidates die.

05

Evolve

Survivors rewrite the system that produced them.

02 Mandates

Three books,
one engine.

Each book decays. The engine is what replaces it.

01

Arbitrage

We take short-lived dislocations where prices disagree with themselves.

02

Prediction

Predictive models carry what arbitrage cannot, and they are replaced as their edge decays.

03

Event agents

Frontier models read the event and act before the reading is priced.

Experiment ledger
internal
Run Question State
arb-041 Edge survival after costs running
event-118 Signal latency against hit rate retained
exp-023 Feature search, overfit or not discarded
Illustrative. Discarded runs stay on the record.

Every result is reproducible from its specification, including the failures.

03 People

Engineers and
theorists, one log.

Half of us build systems and half of us build models, and everyone reads the same experiment log in the morning. We hire for taste in problems rather than for pedigree.

Low-latency systems Agentic engineering Statistical theory Microstructure

The edge is the loop, not the idea.

We speak with allocators and with researchers on the same problem.

compliance.enlitech@gmail.com